Everyone Has Had Excel for Thirty Years. Most People Still Can't Build With It. AI Is Going the Same Way.
Everyone Has Had Excel for Thirty Years. Most People Still Can't Build With It. AI Is Going the Same Way.
AI tools are now free, or close to it, for almost everyone. The gap between people who get a selfie caricature out of them and people who get a working system isn't about access, and it widens with every new model. It's about two habits, and whether your organisation has anyone who has them.
Excel has been sitting on almost every work computer for decades. Nobody in finance is short of access to it.
Yet the range of what people do with it is enormous. In one office, someone adds a column on a calculator and types the total into a cell. Next door, someone has built a timesheet file that takes the roster export, applies the award rules and produces a payroll import with no retyping.
Same software. Same licence. Wildly different results.
AI tools are following the same path, only faster. Pew Research Center's 2026 survey found about half of US adults now use AI chatbots, mostly to look things up. Plenty of the rest is fun: in February, asking ChatGPT for a caricature of yourself and your job became what Creative Bloq called the first big social media craze of the year. Meanwhile a much smaller group is using the same tools to build apps, start businesses and automate whole processes.
Nothing wrong with the caricatures. But why such different outcomes, and why does the difference matter more every few months?
|
42%
Of US adults use chatbots to search for information, the top use. 24% make or edit images or video (Pew, February 2026 survey)
|
1 in 5
US adults who say they're highly confident using chatbots, though about half use them (Pew, 2026)
|
|
10%
Higher conversation success rate for people who had used Claude for six months or more, not explained by task choice or country (Anthropic Economic Index, March 2026)
|
7th
Australia's global rank for per-capita Claude use, at more than four times what its working-age population would predict (Anthropic, March 2026)
|
The tool was never the scarce part
Most households own a cordless drill. In one garage it hangs pictures. In another it has built a deck and a cubby house. Nobody would say the second owner had a better drill.
Excel capability tends to sit on a ladder, and the rungs are easy to recognise:
| Rung | In Excel | In AI tools |
| 1 | Typing numbers into cells | Trying it for fun: images, jokes, a party invitation |
| 2 | SUM, simple formulas, some formatting | Asking questions instead of searching |
| 3 | Lookups, pivot tables, a tidy model | Drafting and polishing your own writing |
| 4 | Power Query, macros, data pulled from other systems | Working a real problem through several rounds, checking and correcting as you go |
| 5 | A small system that runs a process end to end | Building a tool or workflow that keeps working when you close the chat |
The usage data suggests most people are on rungs one and two. In Pew's survey, searching for information is the most common use, nearly a quarter of adults use chatbots to make images or video, and only about one in five say they're highly confident using them. Those are useful, and save real time. But they aren't where the compounding happens. The person on rung five isn't saving ten minutes on an email. They've removed a whole task from the week, permanently. (On Tuesday I showed what that looked like on one of my own projects.)
What actually separates the rungs: two habits
In my view, the gap comes down to two things, and neither of them is technical.
The first is problem solving, in the plain sense: breaking a messy situation into steps you can describe. A good home cook doesn't see "dinner". They see what gets prepped first, what cooks while something else rests, and what goes in last. The person who built the roster-to-payroll file usually didn't start out knowing Power Query. They started with a task that annoyed them every fortnight, wrote down the steps, and went looking for the feature that did each one.
AI rewards exactly that, because it can only help with the problem you can frame. "Fix my month-end" gets generic advice. "Here are the five steps where our month-end waits on someone; which two could be checked automatically?" gets you somewhere.
The second is curiosity: the reflex to ask "is there a better way?" and then actually try it. It's the person who asks why the output was wrong instead of closing the window.
The evidence is early and indirect, but consistent with this. Anthropic's March 2026 Economic Index report found that people who had used Claude for six months or more had about a 10% higher success rate, used it more for demanding work and iterated more. Anthropic is careful about the cause: sophisticated users may simply have arrived first, or it may be learning by doing. Either way, the curious turned up early and the ones who stayed got better. As we covered last month, that kind of growth lasts longer than the output gains.
Australia starts well: Anthropic describes Australian use as more collaborative and less delegated than in many countries, which is where both habits do their work.
Why this is a leadership question, not an IT one
For a CEO or GM, the uncomfortable part is that buying licences doesn't move anyone up the ladder. Organisations commonly roll out a subscription, watch usage climb, and find six months later that little about the work has changed. Everyone has the tool. Nobody has redesigned anything with it.
Worse, a rung-two user can now produce a polished forecast that looks like rung five, and nobody can tell whether it's right. That's AI output passed up the line without an owner.
Where a fractional CFO fits
Most finance teams of two to four people don't have anyone on rung four or five. That isn't a criticism: they were hired to process, reconcile and pay, and they're flat out. Nobody is paid to ask which part of the week shouldn't exist.
Hiring that person full time is sometimes right, as in when in-house is the better call. Often, though, the workload doesn't justify a full-time senior hire, and the gap sits above the bookkeeper (last week's self-check).
A good fractional CFO brings two things to that gap: knowing what a sound month-end, cash view and management reporting pack look like, and having built enough to know which parts can be automated safely. The goal shouldn't be a clever system only the consultant understands. It should be a team a rung or two higher, with the process written down so it survives a resignation.
If you already have someone in-house with both habits, give them the time and the mandate. They're worth more than any tool.
Three questions to ask about your own team
1. Who do people go to when the spreadsheet breaks? That's your rung-four candidate, probably doing it on top of a full job.
2. What's the weekly task everyone complains about, and has anyone tried to fix it? "We've always done it that way" shows you the problem and the missing habit.
3. When someone last used AI on a work problem, did they come back with an answer or a better question? A better question means someone is climbing. A pasted, unchecked answer means someone is standing still.
The same lesson, on a faster clock
Excel didn't make anyone a good analyst. It made good analysts much faster, and it made the gap between them and everyone else much more visible. AI is doing the same thing, on a bigger scale and at a lower price.
With one difference. Excel's ceiling moved slowly. AI's moves every few months, and each move helps the person who can frame a problem far more than the person who only asks questions. A better model makes a rung-two search answer a little better. It lets a rung-five builder do something that wasn't possible last quarter. So the gap grows with every release.
It also means the opportunity has a shelf life. Right now, one person with a clear idea and a couple of subscriptions can build the tool they've always wanted, test the side business they kept putting off, or remove the part of the week they've hated for years, often before competitors start. That advantage comes from being early, and early doesn't last. Once building with AI is ordinary, it earns what any ordinary skill earns. In my view, the window is open now. It won't stay open.
So make the caricature. I won't judge. Just don't confuse having the tool with having the skill. The skill is still problem solving and curiosity, and those, unlike the tools, don't come with a free tier.
Note: The scenarios in this post are based on real experiences — mine and those shared by colleagues across the sector. Details might have been changed and modified slightly to protect confidentiality, and mostly used 1st person perspective for convenience.
Which rung is your finance team on?
The free Finance Function Health Check is a short, no-obligation review of who owns each part of your finance function, where the manual work sits, and which parts could move up a rung safely. PFL provides senior-level outsourced finance, management reporting, and AI automation for Australian NFP, NDIS, and SME organisations.
Book your free health check →- Pew Research Center — Americans and AI 2026: Chatbots, Smart Devices and Views on Impact (17 June 2026; survey 17–23 February 2026)
- Pew Research Center — What do Americans think AI is? (17 June 2026)
- Creative Bloq — This viral AI caricature trend is everywhere (9 February 2026)
- Anthropic — Anthropic Economic Index report: Learning curves (24 March 2026)
- Anthropic — How Australia Uses Claude: Findings from the Anthropic Economic Index (31 March 2026)
Comments
Post a Comment